fix(wrxlib+ci): default WRX_RUN_OK=1, remove obsolete env-gate (0 skip)#166
Merged
Conversation
The historical libwrxapi.so SEGV inside wrcalpwr -> grd1d was fixed in PR #123 by setenv'ing WRX_NO_GRAPHICS=1 inside wrx_api_init. The opt-in gate in test_wrxlib.py is now stale; convert to the default-on pattern already used by sister test files (test_equivalence.py, test_sweep.py, test_property_boundary.py): WRX_RUN_OK=0 still works as a kill switch. Result: test_run_and_get_state runs by default. wrxlib subdir now reports 34 pass / 0 skip (was 33 pass / 1 skip). Note: CI workflow comment at .github/workflows/python-tests.yml:229-235 still warns "Don't default WRX_RUN_OK on" — that comment was written before PR #123 landed; the SEGV root cause is now fixed. If CI surfaces a regression, revisit the comment alongside this change. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The original "Don't default WRX_RUN_OK on" CI rationale was an SIGABRT-avoidance gate predating PR #123 (wrx_api_init sets WRX_NO_GRAPHICS=1 to suppress libgrf::grd1d) and PR #163 (eq EQFINI rearms eq_bpsd_init_flag SAVE so suite-level reinit cycles stop SEGV'ing). With both fixes in place the gate is obsolete and CI should exercise the full wrx_run() surface — leaving it gated would let regressions in wrx run-path code reach develop without ever being exercised by CI. This is the CI half of the gate retirement; the test-side change (test_wrxlib.py defaulting RUN_OK=on) was the previous commit on this branch. Codex review: no findings. feature-dev review: dropped phantom WRX_REINIT_OK from the env block (no Python code reads it; the gate was retired at the Fortran level per wrx/wrcomm.f90:214). README doc-drift cleanup deferred to follow-up task. Verified locally: - pytest python/wrxlib/ : 34 passed, 0 skipped - (CI behavior change is the actual delivery — local already defaults RUN_OK=on after the previous commit) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Owner
Author
|
@cursor review |
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit 738316e. Configure here.
This was referenced Apr 23, 2026
k-yoshimi
added a commit
that referenced
this pull request
Apr 23, 2026
Align wrxlib/README.md, wrx_mcp/README.md + server.py, and the 3 wrxlib/examples/*.py docstrings with the current state of the WRX_RUN_OK gate: * PR #123 fixed the libgrf::grd1d SEGV via WRX_NO_GRAPHICS=1 setenv in wrx_api_init, so wrx_run no longer crashes through dlopen. * PR #163 added EQFINI rearm so reinit cycles are also safe. * PR #166 flipped the test-side WRX_RUN_OK gate to default-on (os.environ.get("WRX_RUN_OK", "1") != "0"). The wrxlib library itself has no runtime gate — .run() dispatches unconditionally. The MCP-level gate in wrx_mcp/server.py is retained as a defence-in-depth / explicit opt-in safeguard so an LLM does not accidentally burn minutes on a ray-tracing run without the user authorising it; the gate error message, module docstring, run / run_and_get_state tool docstrings, and --help text were updated to reflect this rationale instead of the obsolete "SEGV prevention" story. No behavioural change — docs + gate message strings only. wrx_mcp pytest: 37/37 PASS under --forked --timeout=120. Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
k-yoshimi
added a commit
that referenced
this pull request
Apr 27, 2026
…s + ray helper
Extends the tr/eq applications pattern to wrx (extended ray tracing).
Three samples adapted to wrx's per-ray array indexing:
1. safe_run + ray-launch sanity report — pure-Python preflight checks
each ray's RPI/RF/ANGT against [RR-(RA+RB), RR+(RA+RB)] etc.; on
WrxlibRunError it prints which ray is the suspect (since geometry-
error retry is futile).
2. fan_sweep via _apply_rays — fan_sweep([8,10,12]) returns pwr_tot,
pwr_nsa (per-species sum), and pwr_nsa_nray (per-ray x per-species
2D matrix).
3. auto_setup + hand-rolled _hand_validate + DEVICE_PRESETS[
"ITER_LHCD_2ray"] — preset splits {shape, rays}; validate enforces
NSMAX>=1, non-empty rays, REQUIRED_RAY_KEYS = ("RF","RPI","ZPI",
"ANGT") per ray. NRAYMAX = len(rays) auto-derived.
The _apply_rays(wrx, rays) helper takes [{"RF":170e3, "RPI":8.0,
"ZPI":0.5, "ANGT":10.0}, ...], hides 1-origin Fortran indexing,
fills defaults (PHII=0, ANGPHI=0, UU=1, MODEW=1), and writes
RFIN[i]/RPIN[i]/... via set_param.
Reproduced live numeric outputs against libwrxapi.so (ITER 170 GHz,
NSMAX=2 D+e):
Single ray (ANGT=10): pwr_tot=0.7840, pwr_nsa=[0.7840, 0.0]
3-ray fan ANGT=8/10/12: pwr_tot=1.9122
2-ray ITER preset: pwr_tot=1.0073
Bad geometry (RPI=1.0 < 2.0): preflight flags ray 1 then ierr=3
Caveats noted in {note} admonition: historical libgrf::grd1d SEGV
(retired by PR #123), WRX_RUN_OK=1 default since PR #166, and the
pwr_nray[i]==0.0 quirk (use pwr_nsa_nray row sums).
Sphinx wrx-ja build: warning-free (-W --keep-going passes).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
k-yoshimi
added a commit
that referenced
this pull request
Apr 27, 2026
English version of the wrx applications page (safe_run / fan_sweep / auto_setup + _apply_rays helper). Mirrors 3b86091. Numeric outputs (pwr_tot=1.9122, pwr_tot=1.0073, etc.) and PR #123 / PR #166 caveat references preserved verbatim. Verified: sphinx wrx-en build is warning-free. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Two-half retirement of the obsolete
WRX_RUN_OKenv-gate that was once needed to avoid anSIGABRTin libgrf::grd1d. Both root causes are now fixed:9c44546d(test_wrxlib.py edit)RUN_OK = os.environ.get("WRX_RUN_OK", "1") != "0"(default-on, opt-out via=0). Sister filestest_equivalence.py/test_sweep.pyalready used this style738316e2(workflow edit)env: WRX_RUN_OK: "1"on the pytest job so CI explicitly exercises the run pathWhy now
WRX_NO_GRAPHICS=1insidewrx_api_init→ suppresses the libgrf path insidewrcalpwr.f90(the original SEGV)EQFINI→ rearmseq_bpsd_init_flagSAVE so suite-level reinit cycles no longer SEGVwrx_run()from CI.Reviewers
WRX_REINIT_OK: "1"from the env block —WRX_REINIT_OKwas retired at the Fortran level (wrx/wrcomm.f90:214) and is no longer read by any Python code. Fixed in commit738316e2.python/wrxlib/README.md(5 stale references) andpython/mcp-servers/wrx_mcp/README.md(§8) still tell usersWRX_RUN_OK=1is mandatory. Deferred to follow-up task — keeps this PR's scope tight (CI-behaviour-only).WRX_RUN_OK=1enabled and confirmed pass.Test plan
pytest python/wrxlib/: 34 passed, 0 skipped (was 33 + 1 skip;test_run_and_get_statenow runs and passes)pytest python/wrxlib/ python/totlib/ python/mcp-servers/wrx_mcp/: 115 passed, 0 skippedwrx_runwithout SEGV)Fallback if CI breaks
Per the reviewer's risk note: if a different gfortran-13.2 strictness issue surfaces, the symptom will be SIGABRT/SIGSEGV in a
--forkedchild (not a suite-level crash).--maxfail=50provides containment. If it happens: revert this PR, open follow-up to investigate viavalgrind python -m pytest python/wrxlib/....🤖 Generated with Claude Code
Note
Medium Risk
Changes CI/test gating so
wrx_run()paths execute by default, which may surface native-library crashes or flakes in CI even though product code is unchanged.Overview
CI now always exercises
wrx_run()by default. Thepython-testsworkflow setsWRX_RUN_OK=1for the pytest step and drops the old rationale/notes about keeping run tests off in CI.python/wrxlib/tests/test_wrxlib.pyflipsRUN_OKto default-on (WRX_RUN_OK=0opt-out), updates skip messages/docs accordingly, and makes theTestWrxlibRunsuite run wheneverlibwrxapi.sois present unless explicitly disabled.Reviewed by Cursor Bugbot for commit 738316e. Bugbot is set up for automated code reviews on this repo. Configure here.